You are on page 1of 14

Journal of Applied Psychology Copyright 2004 by the American Psychological Association

2004, Vol. 89, No. 3, 483 496 0021-9010/04/$12.00 DOI: 10.1037/0021-9010.89.3.483

The Construct and Criterion Validity of Emotional Intelligence and Its


Potential Utility for Management Studies

Kenneth S. Law Chi-Sum Wong


Hong Kong University of Science and Technology The Chinese University of Hong Kong

Lynda J. Song
Hong Kong University of Science and Technology
and is not to be disseminated broadly.
of its allied publishers.

In this study, the authors reviewed the definition of emotional intelligence (EI) and argued that EI is
conceptually distinct from personality. In Study 1, the authors showed that EI was related to yet distinct
from personality dimensions and that it had incremental predictive power on life satisfaction. The authors
examined the construct validity of self-reports and others ratings of EI using two samples in Study 2.
In a student sample, parents ratings explained additional variance in the students life satisfaction and
feelings of powerlessness after controlling for the Big Five personality dimensions. In the work sample,
peer ratings were found to be significant predictors of job performance ratings provided by supervisors
after controlling for the Big Five personality dimensions. Other implications for future research on EI are
discussed.
This document is copyrighted by the American Psychological Association
of the individual

Emotional intelligence (EI) has been an emerging topic among views of EI, some researchers continue to develop new and revised
social and organizational psychologists in recent years (see, e.g., EI measures (see, e.g., Salovey, Mayer, Caruso, & Lopes, in press;
Fineman, 1993; Mayer & Salovey, 1997; Schutte et al., 2001). Schutte et al., 1998; Stough & Palmer, 2002), whereas others
There have been serious academic discussions on whether EI is an continue to take a conservative position and question the useful-
elusive construct (Davies, Stankov, & Roberts, 1998) on its di- ness of EI for serious scientific research (see, e.g., Petrides &
mensions (Salovey & Mayer, 1990), and on whether it should be Furnham, 2000).
considered as an intelligence dimension (Mayer, Caruso, & The major purposes of this study are threefold. First, we review
This article is intended solely for the personal

Salovey, 2000a; Mayer & Salovey, 1997). Proponents of the EI the definition and domain of the EI construct and argue that when
construct have argued that it is distinct from traditional personality EI is properly defined, it is conceptually distinct from personality
traits and general mental ability and that it is a meaningful con- dimensions. Second, we examine a newly developed EI scale
struct that can be used to explain various psychological and man- under the generally agreed EI definition and demonstrate empiri-
agerial phenomena. These proponents have developed various EI- cally that as measured by this new scale, EI is distinct from
related scales to measure the construct. However, some schol- ars personality dimensions. Third, in addition to conceptual and em-
have voiced strong reservations concerning the reliability and pirical distinctiveness, we also try to establish the predictive va-
validity of these scales. For example, Davies et al. (1988) reviewed lidity of EI in social and organizational settings. We use self- and
all of the existing EI-related scales and demonstrated, through others ratings of EI to investigate its construct validity using
exploratory factor analyses, that most of the scales had salient multitrait-multimethod (MTMM) analyses. Furthermore, we illus-
cross-loadings on personality dimensions. However, Wong and trate the incremental predictive power of others ratings of EI on
Law (2002) reexamined the definition and domain of the EI psychological and work outcomes, over and above personality
construct and developed a new EI scale. Using this new EI scale, dimensions, using a series of hierarchical regression analyses.
they showed that on top of general mental abilities (GMAs), EI
was a good predictor of job performance. Given these opposing
Definition and Domain of EI
Psychology and management researchers were interested in
Kenneth S. Law and Lynda J. Song, Department of Management of studying human emotions long before the construct of EI was
Organizations, Hong Kong University of Science and Technology, Hong proposed. Salovey and Mayer (1990) were among the earliest to
Kong; Chi-Sum Wong, Department of Management, The Chinese Univer- suggest the name emotional intelligence to refer to the ability of a
sity of Hong Kong, Hong Kong.
person to deal with his or her emotions. They defined EI as the
The work described in this article was supported by a Grant
CUHK4038/00H from the Research Grants Council of the Hong Kong
subset of social intelligence that involves the ability to monitor
Special Administrative Region (Project No. CUHK4038/00H). ones own and others feelings and emotions, to discriminate
Correspondence concerning this article should be addressed to Kenneth among them and to use this information to guide ones thinking
S. Law, Department of Management of Organizations, Hong Kong Uni- and actions (p. 189). Despite this early definition of EI, there has
versity of Science and Technology, Clear Water Bay Road, Hong Kong. been confusion regarding the exact meaning and domain of this
E-mail: mnlaw@ust.hk construct. During the early stage of the development of the EI

483
484 LAW, WONG, AND SONG

construct, different researchers used slightly different definitions recovery from psychological distress. A person with high ability in
of EI, which has led to some variations in the domain of the this area would be able to return quickly to normal psychological
construct. As Mayer et al. (2000a) commented, some of these states after rejoicing or being upset. Such a person would also have
alternative conceptions of emotional intelligence [italics added] better control of his or her emotions and would be less likely to
include not only emotion and intelligence per se, but also motiva- lose his or her temper.
tion, non-ability dispositions and traits, and global personal and 4. Use of emotion to facilitate performance. This relates to the
social functioning (p. 268). The Bar-On Emotional Quotient ability of a person to make use of his or her emotions by directing
Inventory (EQi; Bar-On, 1997) is a well-known EI scale that them toward constructive activities and personal performance. A
belongs to this category. person who is highly capable in this dimension would be able to
Among various academics who have contributed to the devel- encourage him- or herself to do better continuously. He or she
opment of the EI construct, two groups of scholars have been of would also be able to direct his or her emotions in positive and
prime importance. They are Davies et al. (1998) and Salovey and productive directions.
Mayer (1990). In 1998, Davies et al. qualitatively summarized the We use Davies et al.s (1998) definition of EI because it is more
EI literature and developed from it a four-dimensional definition of representative of the entire EI literature. Davies et al.s review
and is not to be disseminated broadly.

considered Mayer and Saloveys (1997; Salovey & Mayer, 1990)


of its allied publishers.

EI. However, they did not develop any measure of EI. Instead, they
used earlier work on EI and a group of EI-related measures to definition of EI. In addition, Davies et al.s definition of EI is, in
show that these measures loaded on the same factors as the Big fact, quite similar to that of Mayer and Salovey and also matches
Five personality dimensions. On the basis of these cross-loadings well with Ciarrochi et al.s (2000) summary of the four basic areas
in a series of exploratory factors analyses, Davies et al. concluded of EI. Davies et al.s definition of the dimensions of EI allows us
that EI was elusive as a construct. Ironically, while building up the to focus on the nature and characteristics of the EI construct.
foundation of EI by drawing a four-dimensional definition of EI
from the literature, they used early EI scales that were not based on Nature of EI
this four-dimensional definition and concluded that EI was an
This document is copyrighted by the American Psychological Association

elusive construct. EI has its roots in the concept of social intelligence first iden-
Two years later, Mayer, Caruso, and Salovey (2000b) used a tified by Thorndike in 1920. Thorndike (1920) defined social
slightly different definition of EI and developed the Multifactor intelligence as the ability to understand and manage men and
of the individual

Emotional Intelligence Scale. Mayer et al. followed the conceptu- women, boys and girlsto act wisely in human relations (p. 228).
alization developed by Salovey and Mayer (1990; Mayer & Following Thorndikes ideas, Gardner (1993) included interper-
Salovey, 1997) and defined EI as a set of interrelated skills that can sonal and intrapersonal intelligences in his theory of multiple
be classified within the following four dimensions: intelligences. According to Gardner, social intelligence, which is
one among seven intelligence domains, comprises an individuals
the ability to perceive accurately, appraise, and express emotion; the interpersonal and intrapersonal intelligences. Intrapersonal intelli-
ability to access and/or generate feelings when they facilitate thought;
gence relates to ones ability to deal with oneself and to symbol-
This article is intended solely for the personal

the ability to understand emotion and emotional knowledge; and the


ability to regulate emotions to promote emotional and intellectual
ize complex and highly differentiated sets of feelings (p. 239)
growth. (Mayer & Salovey, 1997, p. 10) within the self. Interpersonal intelligence relates to ones ability to
deal with others and to notice and make distinctions among other
Although the definitions of EI used by Davies et al. (1998) and individuals and, in particular, among their moods, temperaments,
Mayer et al. (2000b) were not identical, the differences in the motivations and intentions (p. 239). EI can be viewed as a
definitions were minor. As Ciarrochi, Chan, and Caputi (2000) combination of the intrapersonal and interpersonal intelligence of
commented in a review of the EI literature, while the definitions an individual.
of EI are often varied for different researchers, they nevertheless Given its roots in social intelligence, Mayer et al. (2000a)
tend to be complementary rather than contradictory (p. 540). argued that the conceptual background of EI met traditional stan-
They further pointed out that in general, the various measures (of dards for intelligence measures. They highlighted three criteria that
EI) cover [italics added] . . . four distinct areas: emotion percep- qualify EI as a facet of intelligence. These are conceptual, corre-
tion, regulation, understanding, and utilization (p. 540). lational, and developmental criteria. The conceptual criterion re-
In this study, we use the four-dimensional definition of EI quires that EI reflect mental abilities instead of preferred ways of
developed by Davies et al. (1998). These four EI dimensions are as behaving. Intelligence refers to the ability of a person. Gardner
follows: (1993) defined intelligence as the ability [italics added] to solve
1. Appraisal and expression of emotion in oneself. This relates problems, or to fashion products, that are valued in one or more
to an individuals ability to understand his or her deep emotions cultural or community settings (p. 7). To fulfill the conceptual
and to be able to express emotions naturally. People who have criterion and label EI as an intelligence facet, one must provide
good ability in this area will sense and acknowledge their emotions evidence that EI is not a personality trait or a preferred way of
better than most people. behaving but is itself a set of abilities. We argue conceptually that
2. Appraisal and recognition of emotion in others. This relates EI is defined as a set of abilities on the basis of its theoretical
to an individuals ability to perceive and understand the emotions definition. We also cite empirical evidence that EI has been shown
of the people around them. People who rate highly in this ability to be distinct from personality dimensions as follows.
will be very sensitive to the emotions of others as well as able to As defined earlier, EI consists of a set of abilities that a person
predict others emotional responses. uses to understand, regulate, and make use of his or her emotions.
3. Regulation of emotion in oneself. This relates to the ability of Emotional understanding, regulation, and utilization reflect the
a person to regulate his or her emotions, enabling a more rapid capability of a person to manage his or her emotions. Some people
CONSTRUCT AND CRITERION VALIDITY OF EI 485

have higher competence to do so; some have less competence. As or WLEIS1 for easy reference) following the four-dimensional
a particular type of competence, EI is different from personality definition of EI as introduced by Davies et al. Wong and Law used
traits that reflect tendencies to think, feel, and behave in certain both exploratory and confirmatory factor analyses (CFAs) in the
ways. In this competence sense, the four-dimensional definition scale development and validation process to show that EI as
qualifies EI as abilities and, therefore, as one possible facet of measured by the WLEIS, was distinct from the Big Five person-
intelligence (see, e.g., Mayer & Salovey, 1997; Mayer et al., ality dimensions. To examine the overall usefulness of the EI
(2000a); Salovey & Mayer, 1990). There is also empirical evi- construct, we used CFA in this study to compare the distinctive-
dence that EI abilities are distinct from personality traits. When ness of the WLEIS and various EI-related scales included in the
developing the Multifactor Emotional Intelligence Scale, Mayer Davies et al. study from the Big Five personality dimensions. On
and Salovey (1997) provided some preliminary data to show that the basis of the data collected, we found that the WLEIS was
EI was different from personality traits. Wong and Law (2002) distinct from the Big Five personality dimensions, although they
also showed in multiple samples that EI as measured under Davies were moderately correlated because of conceptual overlaps.
et al.s (1998) four-dimensional definition, was empirically dis- Because EI is conceptualized as an intelligence facet, it is an
tinct from the Big Five personality dimensions. ability measure instead of a personality measure. This leads to our
and is not to be disseminated broadly.

Mayer et al.s (2000a) second criterion is the correlational


of its allied publishers.

first hypothesis, which is the core focus of Study 1:


criterion. If EI is an intelligence facet, it should represent a set of
correlated abilities that are similar to, but distinct from, mental Hypothesis 1: EI is distinct from, but correlated with, the Big
abilities (p. 270). In other words, if EI is an intelligence facet, it Five personality dimensions.
should correlate with other intelligence facets, such as GMAs.
These correlations cannot be too high (discriminant validity), nor The construct validity of the WLEIS was further investigated by
can they be too low (convergent validity). Both Mayer et al. and the MTMM method with self- and others ratings of EI. In Study
Wong and Law (2002) showed empirically that the dimensions of 2, we deliberately chose two criterion variables from the EI liter-
EI were moderately correlated among themselves but only mildly ature, life satisfaction and powerlessness, which should have pos-
itive and negative correlations with EI, respectively. Life satisfac-
This document is copyrighted by the American Psychological Association

correlated with general mental abilities. There is, therefore, some


empirical evidence that EI meets the correlational criterion of an tion was used as a criterion variable because proponents of EI have
intelligence facet. argued that theoretically, life satisfaction should be positively
related to EI. Intrapersonal emotional recognition and management
of the individual

Mayer et al.s (2000a) third argument that EI could be concep-


tualized as an intelligence facet is that intelligence should be helps an individual deal with his or her emotions. A person with
developmental in nature. The verbal ability of a person, for exam- high EI should be able to recognize his or her emotions, to regulate
ple, should increase as one becomes more mature. Wong, Wong, those emotions, and to use them to facilitate performance. As a
and Law (2002) found that EI is positively correlated with age result, this person should be happier as a whole in life. Several
among incumbents of six different types of jobs. Mayer et al. also empirical studies have provided evidence of this positive relation-
ship (e.g., Wong & Law, 2002; Wong et al., 2002). Powerlessness
This article is intended solely for the personal

showed with a series of studies, that EI increased with age and


experience, which qualifies it as an ability rather than a personality refers to the extent to which one regards ones experiences as
trait. being fatalistically determined, in contrast with their being under
ones control (Pearlin & Schooler, 1978). Specifically, powerless-
ness is similar to a temporal variable showing ones state of mind
Discriminant and Predictive Validity of EI toward ones control over life problems and experiences. A person
On the basis of the above discussion, it is clear that the defini- with high EI would be less affected by his or her emotions, be able
tion and domain of EI have been gradually established. It also to direct emotions in a positive direction, and have lower chances
seems clear, however, that further and stronger evidence that EI of feeling depressed. EI also includes a dimension of understaning
possesses the three basic characteristics of an intelligence facet as others emotions. A person with high EI would also be able to
suggested by Mayer et al. (2000a) is needed. To establish the interpret others moods correctly and therefore have a higher
construct validity of EI, we need to demonstrate its partial (or chance of forming close relationships and getting social support in
complete) independence from other, seemingly analogous, con- general. As a result, this person should be less likely to feel
cepts (Davies et al., 1998, p. 989). As mentioned earlier, Davies powerless. There is some evidence in the literature that feelings of
et al. (1998) factor analyzed some existing EI-related scales and powerlessness are related to negative emotions such as sadness and
personality measures in multiple samples and concluded that EI fear (e.g., Roseman, Dhawan, Rettek, & Naidu, 1995). These ideas
was elusive. lead to our second and third hypotheses:
One possible reason for Davies et al.s (1998) finding is that the
EI-related scales used in their investigations were early scales Hypothesis 2: EI is positively associated with life satisfaction.
related to EI. During that time, a definition of EI had not been
Hypothesis 3: EI is negatively associated with feelings of
systematically developed. Davies et al. also admitted that the
powerlessness.
scales they identified were only EI-related scales. These scales
were not developed according to the four-dimensional view then Finally, EI will be a useful construct for management studies if
proposed by Davies et al. In fact, some earlier scales had not even it has predictive power for job performance. It is intuitive that EI
been identified as EI scales by their developers. The use of these
scales in identifying EI may, therefore, be questionable and unfair.
In response to the findings by Davies et al., Wong and Law (2002) 1
The 16 items of the WLEIS and a short description of the scale
developed a new EI scale (labeled as the Wong and Law EI Scale, development process are presented in the Appendix.
486 LAW, WONG, AND SONG

would affect job performance. Understanding and regulation of distinctiveness of the EI construct using CFAs. The results of
ones emotions as well as understanding of others emotions are Study 1 were threefold. First, a series of CFAs on Sample 1
the core factors affecting intrapersonal well-being and interper- illustrated that the EI-related scales used by Davies et al. are not
sonal relations. In a negative sense, a person who is not sensitive psychometrically sound. Second, CFAs of the new WLEIS, devel-
to his or her own emotions and who is unable to regulate his or her oped on the basis of Davies et al.s four-dimensional view, showed
emotions would have problems interacting with others. Because that EI is distinct from the Big Five personality dimensions. This
interpersonal interactions are a basic component of many jobs, it is result is cross-validated with Sample 2. Third, the WLEIS ex-
reasonable to hypothesize that EI would affect job performance in plained incremental variance in life satisfaction over and above
general. On the positive side, the use of the emotion dimension of that accounted for by the Big Five personality dimensions in both
EI describes ones ability to direct emotions to performance. A Samples 1 and 2.
person with high EI would be able to direct positive emotions to Data for the first sample in Study 1 were collected from
high performance and redirect negative emotions to generate con- university students in two stages. In the first stage, we included
structive performance goals. As Mayer and Salovey (1997) sum- only three of the five self-report EI-related scales that were used
marized, using the emotions as one basis for thinking, and think- in the Davies et al. (1998) study, because two of the EI-related
and is not to be disseminated broadly.

ing with emotions themselves, may be related to important social


of its allied publishers.

scales were not made available to us. After all five EI-related
competencies and adaptive behavior (p. 22). scales were made available, we collected data using all of the
Theoretically, if we follow the social exchange framework scales from an additional sample of students. The data collected
(Blau, 1964), in which organizations use different types of ex- during these two stages were combined to form Sample 1 of
changes to facilitate performance, and social exchange is one Study 1. In both stages of data collection, we also used the
important type of exchange in organizations, then we know that WLEIS. Given the positive results from Sample 1, which con-
factors that facilitate social exchanges improve employee perfor- firmed our arguments, we cross-validated the results with an-
mance. Social exchange involves interactions. As we argued other sample of university students. This cross-validation sam-
above, emotional understanding, regulation, and utilization would ple formed Sample 2 of Study 1.
This document is copyrighted by the American Psychological Association

help to cultivate positive social interactions and exchanges in an


organization and, as a result, facilitate employee performance.
Method
There have also been similar arguments used in the previous
of the individual

literature to justify an EIperformance link. On the basis of At- Sample and Procedures
water and Yammarinos (1992) work on self other rating agree-
ments, Sosik and Megerian (1999) argued that ones understanding Respondents in the first stage consisted of 202 undergraduate students at
of ones and others affective ratings (as a result of ones EI) a large university in Hong Kong. The students were asked to make self-
would influence self other rating agreements, which, in turn, assessments on EI-related scales, Big Five personality scales, and life
satisfaction measures in an introductory management course. We included
would influence performance outcomes. Lam and Kirby (2002)
most of the self-report measures that were examined by Davies et al. (1998)
This article is intended solely for the personal

showed that overall, EI explains individual cognitive-based per-


for easy comparison. The EI-related scales that were included at this stage
formance above and beyond the level attributable to general intel- were the Trait Meta-Mood Scale (TMM; Salovey, Mayer, Goldman, Tur-
ligence. Dulewicz and Higgs (2000) showed that EI was predictive vey, & Palfai, 1995), the Emotional Empathy Scale (EES; Mehrabian &
of the career success of 100 managers. Epstein, 1970), and the Toronto Alexithymia Scale (TAS; Bagby, Taylor,
On top of showing a main effect of EI on job performance, & Parker, 1994). After obtaining items for the Emotional Control Ques-
Wong and Law (2002) argued that emotional labor moderates the tionnaire (ECQ; Roger & Najarian, 1989) and the Affective Communica-
EIperformance link and provided preliminary evidence for EIs tion Test (ACT; Friedman, Prince, Riggio, & Dimatteo, 1980), in our
predictive validity on job performance. Emotional labor refers to second stage of data collection, we added these two scales to all of the
the extent to which an employee is required to present an appro- measures that were used in the first stage. The respondents in the second
priate emotion in order to perform a job in an efficient and stage consisted of 216 undergraduate business students from another large
university in Hong Kong. They filled out this survey as a partial fulfillment
effective manner (Hochschild, 1983). Both Lam and Kirby (2002)
of the course requirements for an introductory organizational behavior
and Wong and Law used self-ratings of EI. In the current study, we course. Data collected in these two stages were pooled together to form
used colleagues ratings of EI to predict supervisory ratings of job Sample 1 of Study 1. The sample size for the Big Five personality scales,
performance. This design provided much stronger support for the the TMM, the EES, and the TAS was 418, and the sample size for the ECQ
utility of the EI construct in management studies. The above and ACT scales was 216.
discussion leads to the final hypothesis of this study: Having obtained evidence to confirm our arguments, we collected a
further sample to provide cross-validation data on personality measures,
Hypothesis 4: EI is positively associated with employees job the WLEIS, and life satisfaction. This sample consisted of 314 undergrad-
performance. uate students from the same university as the participants in the first stage
of Sample 1. All questionnaires were administered in English to the Hong
Hypothesis 1 was tested in Samples 1 and 2 of Study 1. Hy- Kong participants.
potheses 2 and 3 were tested with Sample 1 of Study 2. Hypothesis
4 was tested with Sample 2 of Study 2. Measures
EI-related measures. We used the same self-report EI-related mea-
Study 1 sures as did Davies et al. (1998). These included (a) the 30-item TMM,
which measures three EI subscales: Attention, Repair, and Clarity; (b) the
In this study, we collected data that are comparable to the 33-item EES; (c) the 20-item TAS, which measures EI with three sub-
Davies et al. (1998) data, which enabled a more rigorous test of the scales: Difficulty Identifying Feelings, Difficulty Describing Feelings, and
CONSTRUCT AND CRITERION VALIDITY OF EI 487

Externally Oriented Thinking; (d) the 32-item ECQ; which measures four cross-validation sample (Sample 2 of Study 1), we conducted a CFA and
EI-related subscales: Rehearsal, Emotional Inhibition, Benign Control, and a hierarchical regression to cross-validate the results of the WLEIS used in
Aggression Control; and (e) the 13-item ACT.). Sample 1.
Because all of the above EI-related scales did not measure EI on the
basis of Davies et al.s (1998) proposed four-dimensional definition, we
Results
also included the newly developed 16-item WLEIS as a final measure of
EI. The response format of the WLEIS is a 7-point Likert-type scale (1 = Summary Statistics
totally disagree to 7 = totally agree). A sample item from the Self-
Emotions Appraisal is I really understand what I feel. A sample item Descriptive statistics, coefficient alphas, and correlations among
from the Use of Emotion to facilitate performance dimension is I would all the measures for the two samples are presented in Tables 1 and
always encourage myself to try my best. A sample question from the 2. Results in Table 1 indicate that most of the measures have
Regulation of Emotion dimension is I can always calm down quickly acceptable reliability estimates. Table 2 shows that some person-
when I am very angry. A sample item from the Others Emotion Appraisal ality and EI-related dimensions were moderately correlated, al-
is I have good understanding of the emotions of people around me. The though none of the correlations exceeded .50 except for the cor-
full WLEIS scale and a short description of its development process are
relation between the ACT and Extraversion (r = .59).
and is not to be disseminated broadly.

presented in the Appendix.


of its allied publishers.

Personality. In Sample 1, we used the short form of the Neuroticism,


Extraversion, and Openness Personality Inventory to capture the Big Five CFA of the Factorial Structure of EI
personality dimensions (Costa & McCrae, 1985). Each dimension con-
tained 12 items. In Sample 2 (the cross-validation sample), we used the Big Table 3 shows the results when each EI-related measure was
Five adjective scale that was developed by McCrae and Costa (1987). factor analyzed with the Big Five personality dimensions through
Because all the data were collected in class as an exercise on dispositional the use of CFA. We included only one EI-related measure in each
variables, we had limitations on the length of the questionnaires. To be CFA for two reasons. First, the sample size of 216 did not allow us
fully comparable to the Davies et al. (1998) study, we included all five EI- to include more measures without a severe loss of statistical power.
related scales. To limit the length of the questionnaire, we randomly Second, and more important, we could examine the structure and
This document is copyrighted by the American Psychological Association

selected 6 items for each of the Big Five personality dimensions from the goodness of fit of each EI-related measure if we study each one
original 80-item scale, resulting in a 30-item measure. Because all the independently. The first row of Table 3 shows that when the 9
items within each Big Five dimension of the McCrae and Costa scale
TMM indicators (3 indicators for each TMM dimension) were
measure the same construct, random selection of 6 items from each
of the individual

factor analyzed with the 15 Big Five personality indicators (3


dimension would not affect the validity of the scale. Classical measurement
theory argues that fewer items from each dimension might lower the
indicators for each Big Five personality dimension) and eight
reliability of the measures. However, our results showed that reliabilities of latent factors were specified, the model fit indices were very
the Big Five dimensions were comparable to the original 80-item scale in marginal (comparative fit index [CFI] = .88; TuckerLewis Index
the literature (Costa & McCrae, 1985). In fact, the coefficient alphas of the [TLI] = .85; root-mean-square error of approximation [RMSEA]
five personality dimensions were higher than those of the short form of the = .062) The last row under TMM in Table 3 (i.e., the fourth row
NEO-PI in Sample 1, indicating that the reliabilities of these 30 items were
This article is intended solely for the personal

in the table) shows that when a second-order factor was added


acceptable. underlying the 3 TMM factors, the model fit indices dropped
Life satisfaction. The nine-item life satisfaction measure, constructed
significantly (CFI = .83; TLI = .81; RMSEA = .072). These
by A. Campbell, Converse, and Rodgers (1976), was adopted. This is a
results led to two conclusions concerning the TMM scale. First, the
well-established measure of the general satisfaction of an individual toward
his or her life. scale might not have a clear factorial structure distinct from the
Big Five dimensions. Second, there might not be a higher order
factor underlying the three TMM dimensions.
Analyses Table 3 shows that the results of the CFA for many of the EI-
related scales also have quite unsatisfactory fit indices. For
We conducted two sets of analyses of Sample 1. In the first set of example, the fit indices for the ECQ were quite low (equal to or
analyses, we randomly grouped the items to form three indicators for each less than .85) for both the first-order and the second-order models.
EI-related dimension and each personality dimension and used these indi-
The same was true for the ACT (CFI = .85; TLI = .80; RMSEA =
cators as inputs in the CFA. This method of averaging indicators to form
new indicators to reduce the number of observed variables in CFAs is quite
.077). The first-order model fit for the TAS was marginal (CFI =
common in the literature (see, e.g., Mathieu & Farr, 1991; Mathieu, .90; TLI = .88; RMSEA = .055), whereas the second-order model
Hofmann, & Farr, 1993). We used CFA instead of exploratory factor fit (three TAS factors, one second-order TAS factor, and five Big
analysis because the factor structure of all the constructs was known. Our Five personality factors) was quite low (CFI = .86; TLI = .84;
purpose in Study 1 was to confirm the factor structure of the EI and RMSEA = .064). The above analyses led to another possible
personality scales and to study their interfactor correlations to determine explanation of Davies et al.s (1998) results. Although it may still
whether they are distinct constructs. be true that these EI-related measures are cross-loaded with the
In the second set of analyses, we conducted a series of hierarchical Big Five personality dimensions, the cross-loadings may be a
regressions to show the incremental validity of the WLEIS in predicting
result of their poor representation of the EI construct. In other
life satisfaction above and beyond the Big Five personality dimensions. We
words, Davies et al.s findings might be due to the construct
first entered the personality dimensions as control variables for predicting
life satisfaction. The four dimensions of the WLEIS scale were then added
validity of the EI-related scales that were included in their
as the next set of predictors. If the change in model R2 after adding the investigation.
WLEIS were significant, EI would have incremental explanatory power in The last two rows of Table 3 show the CFA results when the
predicting life satisfaction above and beyond the personality dimensions. WLEIS was analyzed with the Big Five dimensions. In Sample 1,
This would be considered an additional piece of evidence to demonstrate both the first-order model (CFI = .92; TLI = .91; RMSEA =
that EI and the personality dimensions are distinct constructs. On the .053) and the second-order model (CFI = .91; TLI = .90;
488 LAW, WONG, AND SONG

Table 1
Descriptive Statistics and Coefficient Alphas Among All Measures in Study 1

Sample 1 Sample 2
No. of
Variables items M SD a M SD a

Neuroticism 12/6 3.10 0.60 .81 3.50 0.77 .75


Extraversion 12/6 3.28 0.46 .70 3.84 0.76 .79
Openness 12/6 3.19 0.41 .57 3.72 0.64 .61
Agreeableness 12/6 3.29 0.41 .62 4.28 0.60 .67
Conscientiousness 12/6 3.30 0.45 .67 4.00 0.71 .70
TMM1 (Repair) 6 3.36 0.65 .70
TMM2 (Attention) 13 3.52 0.46 .73
TMM3 (Clarity) 11 3.13 0.45 .66
EES (Emotional Empathy) 33 3.30 0.31 .74
and is not to be disseminated broadly.

TAS1 (Difficulty Identifying Feelings) 7 2.72 0.68 .79


of its allied publishers.

TAS2 (Difficulty Describing Feelings) 5 2.90 0.68 .69


TAS3 (Externally Oriented Thinking) 8 2.74 0.44 .46
ECQ1 (Rehearsal) 8 2.86 0.68 .76
ECQ2 (Emotional Inhibition) 8 2.67 0.60 .69
ECQ3 (Benign Control) 8 2.94 0.55 .64
ECQ4 (Aggression Control) 8 3.05 0.44 .49
ACT (Affective communication) 13 3.06 0.46 .68
WLEIS1 (Self-Emotions Appraisal) 4 4.01 1.05 .89 4.84 1.09 .90
WLEIS2 (Regulation of Emotion) 4 3.78 1.12 .89 4.27 1.20 .89
WLEIS3 (Use of Emotion) 4 4.09 0.92 .80 4.57 0.99 .79
WLEIS4 (Others-Emotions Appraisal) 4 4.15 0.96 .89 4.60 1.15 .93
This document is copyrighted by the American Psychological Association

Life Satisfaction 9 4.92 0.89 .90 4.80 0.89 .90

Note. N = 418 for all variables in Sample 1 except for the ACT and the ECQ, for which N = 216; N = 314
for Sample 2. a = the coefficient alpha of the scale or dimension; 12/6 = 12 items were used for Sample 1, and
of the individual

6 were used for Sample 2; TMM1, TMM2, and TMM3 = the three dimensions of Trait Meta-Mood Scale;
EES = Emotional Empathy Scale; TAS1, TAS2, and TAS3 = the three dimensions of the Toronto Alexithymia
Scale; ECQ1, ECQ2, ECQ3, and ECQ4 = the four dimensions of the Emotional Control Questionnaire; ACT =
Affective Communication Test; WLEIS1, WLEIS2, WLEIS3, and WLEIS4 = the four dimensions of the Wong
and Law Emotional Intelligence Scale.
This article is intended solely for the personal

RMSEA = .057) showed good model fit. Again, in the cross- its four dimensions. In addition, when properly defined, EI is
validation sample (Sample 2), the first-order model (CFI = .92; related to, yet different from, the Big Five personality dimensions.
TLI = .90; RMSEA = .056) showed good model fit and the Hypothesis 1 was supported by the data.
second-order model (CFI = .90; TLI = .89; RMSEA = .060)
showed marginal fit.
These results support two important conclusions regarding the Hierarchical Regression Analyses
four-dimensional EI construct. First, the four EI dimensions, as
measured by the WLEIS, were distinct from the Big Five dimen- Table 4 shows the results of the hierarchical regression for
sions. Although some EI dimensions were moderately correlated both Sample 1 and the cross-validation sample (Sample 2). The
results were quite similar in both of these samples. Neuroticism,
with Neuroticism (r = .44 and .52 in Samples 1 and 2, respec-
Extraversion, and Agreeableness were the three major personal-
tively) and Conscientiousness (r = .59 in both samples), EI is
ity traits affecting life satisfaction. When the four EI dimen-
different from these personality dimensions. The correlations are
sions were added to the regression model, the increases in the
expected because EI is conceptually related to these dimensions.
For example, neuroticism is a general tendency to overreact to model multiple correlation squared were significant ( p < .01),
negative stimuli from the environment. Although neuroticism is although the absolute magnitude was not large (AR2 = .05 and
conceptually distinct from EI, which reflects ones ability to un- .06 for Sample 1 and Sample 2, respectively). The 5% 6% addi-
derstand, regulate, and use ones emotions, it is very likely that a tional variance accounted for in life satisfaction is understand-
person with low EI would be a neurotic person as well. We able, because as we saw in Table 2, EI is moderately correlated
therefore see the moderate correlations between EI and some with the Big Five personality dimensions. In addition, life satis-
personality dimensions, as expected. However, results of the CFA faction was moderately correlated with the personality dimen-
clearly showed that EI is distinct from the Big Five personality sions; the Pearson correlations between life satisfaction and
dimensions. The second conclusion regarding the WLEIS, as il- Neuroticism, Extraversion, Openness, Agreeableness, and Consci-
lustrated in Table 3, is that a meaningful overall construct under- entiousness were .48, .43, .18, .26, and .33, respectively. There was
lies the four EI dimensions. The second-order model for the limited room for additional variance to be accounted for by EI
WLEIS generated acceptable fit indices in both the first sample when the Big Five personality dimensions were in the model. As
and the cross-validation sample. On the basis of the above analy- a result, we interpreted the additional 5% to be of reasonable
ses, we concluded that EI is an overall latent construct underlying practical significance.
CONSTRUCT AND CRITERION VALIDITY OF EI 489

Table 2
Correlations Among Measures in Sample 1 and Sample 2 of Study 1

Variable 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22
1. NEURO .01 .13 .05 .24 .26 .45 .24 .07 .35
2. EXTRA .38 .39 .24 .22 .23 .01 .22 .19 .27
3. OPEN .01 .13 .00 .10 .13 .13 .16 .20 .03
4. AGREE .20 .23 .10 .21 .19 .12 .05 .08 .20
5. CON .29 .25 .03 .30 .28 .26 .36 .21 .13
6. TMM1 .48 .47 .15 .31 .26
7. TMM2 .09 .18 .25 .22 .05 .09
8. TMM3 .36 .17 .14 .16 .37 .29 .16
9. EES .25 .11 .28 .35 .05 .08 .40 .03
10. TAS1 .45 .19 .04 .37 .35 .33 .17 .53 .09
11. TAS2 .36 .24 .08 .20 .24 .21 .12 .48 .04 .48
.06 .21 .40 .21 .13 .24 .36 .25 .40 .25 .20
and is not to be disseminated broadly.

12. TAS3
.39 .27 .15 .30 .24 .38 .05 .34 .07 .42 .23 .15
of its allied publishers.

13. ECQ1
14. ECQ2 .01 .37 .12 .18 .15 .22 .23 .33 .37 .23 .53 .30 .14
15. ECQ3 .43 .09 .06 .35 .38 .26 .05 .40 .06 .45 .32 .26 .25 .05
16. ECQ4 .01 .12 .12 .34 .11 .10 .00 .02 .09 .10 .08 .14 .12 .05 .05
17. ACT .23 .59 .29 .13 .18 .31 .20 .20 .22 .07 .36 .34 .12 .43 .05 .21
18. WLEIS1 .30 .06 .15 .09 .30 .19 .02 .48 .04 .04 .31 .04 .12 .33 .25 .04 .19 .34 .30 .45 .25
19. WLEIS2 .43 .06 .03 .06 .27 .21 .14 .28 .25 .19 .08 .05 .25 .08 .43 .10 .03 .54 .23 .14 .13
20. WLEIS3 .22 .11 .22 .02 .44 .23 .01 .32 .09 .11 .12 .07 .16 .27 .12 .18 .24 .54 .52 .27 .32
21. WLEIS4 .01 .03 .16 .04 .22 .05 .00 .16 .07 .06 .13 .03 .00 .27 .14 .14 .21 .41 .33 .44 .12
22. LIFESAT .48 .44 .18 .26 .33 .48 .19 .33 .11 .29 .23 .24 .35 .31 .21 .16 .41 .34 .33 .37 .17
This document is copyrighted by the American Psychological Association

Note. Figures on the lower triangle are correlations for Sample 1; numbers on the upper triangle are correlations for Sample 2. N = 418 for all variables
in Sample 1 except for the ACT and the ECQ, for which N = 216; N = 314 for Sample 2. NEURO = Neuroticism; EXTRA = Extraversion; OPEN =
Openness; AGREE = Agreeableness; CON = Conscientiousness; TMM1, TMM2, and TMM3 = the three dimensions of Trait Meta-Mood Scale; EES =
Emotional Empathy Scale; TAS1, TAS2, and TAS3 = the three dimensions of the Toronto Alexithymia Scale; ECQ1, ECQ2, ECQ3, and ECQ4 = the four
of the individual

dimensions of the Emotional Control Questionnaire; ACT = Affective Communication Test; WLEIS1, WLEIS2, WLEIS3, and WLEIS4 = the four
dimensions of the Wong and Law Emotional Intelligence Scale; LIFESAT = Life Satisfaction.

Study 2 personality dimensions. In both cases, common method vari-


ance could be avoided because the dependent variable and the
This article is intended solely for the personal

Our major purpose in Study 1 was to show that with appropriate predictor (i.e., EI) were collected from different sources.
measures, EI is distinct from the Big Five personality dimensions.
EI also accounts for a significant proportion of the variance in life
satisfaction, above and beyond that of the Big Five personality Method
dimensions. However, the self-report nature of the WLEIS limited
the validity of the evidence in Study 1. For example, its explana- Participants
tory power for life satisfaction may be due to common method Sample 1. In this first sample, we collected data by distributing 560
variance (see, e.g., Bagozzi & Yi, 1990; Williams & Brown, sets of questionnaires in two high schools in Anhui province of the
1994). Therefore, we needed additional evidence to demonstrate Peoples Republic of China. All respondents were students in Grade 12 of
the predictive power of EI using non-self-report measures of EI. high school. Each questionnaire set included one self-rating questionnaire
Study 2 addressed this issue by providing evidence of discrimi- to be completed by the students and one parent-rating questionnaire to be
nant and convergent validity, using multiple raters to assess EI. completed by their parents. Students were asked to fill in their self-rating
questionnaire, which assessed their own EI, the Big Five personality
Study 2 consisted of two samples: one of high school students and
dimensions, their life satisfaction, and their perception of powerlessness.
their parents and the other a work sample with self-, peer, and
One parent of each student received the parent questionnaire, which asked
supervisory ratings of EI and job performance. them to evaluate the EI and personality of their son or daughter. Out of the
Study 2 contributed to the criterion validity of EI as well as 560 sets of questionnaires (i.e., 1,120 questionnaires), we received 889
to its construct validity. To demonstrate the construct validity valid responses, a return rate of 79.4%.
of EI, we used multiple assessors (parent and self-ratings of Sample 2. We collected data from a cigarette factory in Anhui prov-
students in Sample 1; self-, peer, and supervisory ratings in ince. One hundred eighty-one sets of questionnaires were distributed to
Sample 2) to measure EI and related constructs and analyzed first-line supervisors in the factory. In each set, there were three question-
the data using the MTMM technique. To demonstrate the cri- naires, sealed in separate envelopes. Each supervisor received one set of
terion validity of EI, we used parents ratings of EI to predict questionnaires in which he or she was asked to rate two of his or her
immediate subordinates (Subordinates A and B). The supervisor was then
students self-ratings of life satisfaction and feelings of power-
instructed to give the two subordinate questionnaires to Subordinates A and
lessness, after controlling for the Big Five personality dimen-
B separately. Subordinate A was required to rate his or her own EI and job
sions. By making use of the work sample in Sample 2, we performance as well as the EI and job performance of his or her peer (i.e.,
showed that peers ratings of an employees EI were a useful Subordinate B), whose name was written on the questionnaire by the
predictor of a supervisors rating of job performance of the supervisor. Similarly, Subordinate B was asked to rate his or her own EI
same employee, after controlling for relevant variables and and job performance as well as that of Subordinate A. All the question-
490 LAW, WONG, AND SONG

Table 3
Results of Confirmatory Factor Analyses for Study 1

Model fit indices Factors

Measure y2 df RMSEA TLI CFI IFI Dimension NEURO EXTRA OPEN AGREE CON

TMM a
M1d = 568.60 224 .062 .85 .88 .88 TMM1 .59 .66 .25 .42 .36
TMM2 .11 .23 .49 .35 .08
TMM3 .46 .28 .15 .30 .54
M2 = 714.72 234 .072 .81 .83 .84 Second order .72 .78 .39 .59 .60
EESa 336.50 120 .066 .86 .89 .89 .33 .16 .54 .49 .08
TASa M1 = 485.79 224 .055 .88 .90 .90 TAS1 .57 .26 .06 .50 .52
TAS2 .46 .32 .09 .27 .36
TAS3 .10 .04 .81 .39 .26
M2 = 609.72 234 .064 .84 .86 .86 Second order .61 .34 .15 .52 .55
M1 = 520.94 .40 .05 .46 .37
and is not to be disseminated broadly.

ECQb 288 .062 .81 .85 .85 ECQ1 .51


.53 .22 .25 .27
of its allied publishers.

ECQ2 .07
ECQ3 .56 .14 .09 .56 .65
ECQ4 .02 .18 .21 .64 .21
M2 = 618.71 305 .071 .76 .79 .80 Second order .92 .50 .18 .87 .95
ACTb 287.60 120 .077 .80 .85 .85 .29 .78 .35 .18 .28
WLEIS (Sample 1)a M1 = 846.41 398 .053 .91 .92 .92 WLEIS1 .37 .11 .12 .11 .42
WLEIS2 .50 .09 .22 .10 .38
WLEIS3 .25 .13 .02 .00 .59
WLEIS4 .00 .04 .01 .06 .29
M2 = 946.96 415 .057 .90 .91 .91 Second order .44 .14 .12 .09 .59
WLEIS (Sample 2)c M1 = 794.07 398 .056 .90 .92 .92 WLEIS1 .31 .28 .19 .23 .36
This document is copyrighted by the American Psychological Association

WLEIS2 .53 .00 .17 .16 .28


WLEIS3 .32 .30 .25 .10 .47
WLEIS4 .08 .21 .27 .09 .25
M2 = 881.43 415 .060 .89 .90 .90 Second order .52 .37 .26 .27 .59
of the individual

Note. RMSEA = root-mean-square error of approximation; TLI = TuckerLewis Index; CFI = comparative fit index; IFI = incremental fit index;
NEURO = Neuroticism; EXTRA = Extraversion; OPEN = Openness; AGREE = Agreeableness; CON = Conscientiousness; TMM1, TMM2, and
TMM3 = the three dimensions of Trait Meta-Mood Scale; EES = Emotional Empathy Scale; TAS1, TAS2, and TAS3 = the three dimensions of the
Toronto Alexithymia Scale; ECQ1, ECQ2, ECQ3, and ECQ4 = the four dimensions of the Emotional Control Questionnaire; ACT = Affective
Communication Test; WLEIS1, WLEIS2, WLEIS3, and WLEIS4 = the four dimensions of the Wong and Law Emotional Intelligence Scale.
a
N = 418. b N = 216. c N = 314. d M1 = the result applicable to the model that specifies only the first order dimensions for the emotional intelligence
This article is intended solely for the personal

measure under investigation. M2 = the result applicable to the model that also specifies a second order latent construct behind the dimensions for the
emotional intelligence measure under investigation.

naires were sealed in envelopes and returned to the human resources Powerlessness in life mastery. This was measured by the 7-item scale
management office of the factory within 2 weeks of dissemination. With constructed by Pearlin and Schooler (1978). The coefficient alpha for the
the enthusiastic support and encouragement of the factory managers, we seven items was .64.
received 165 valid response sets (response rate = 90.7%). Because the Control variables. We collected data on students gender (0 for male,
English standard of students and workers in mainland China is not high, we 1 for female), whether the students mother worked outside the home (0
followed the standard back-translation procedure (Brislin, 1980) to trans- for full-time, 1 for not employed), and the educational level of the
late all the items into Chinese. students father and mother (with 1 to 6 indicating primary school, grade
school, high school, two-year college program, four-year uni- versity
Measures in Sample 1 program, and graduate school) as control variables, respectively.

EI and personality. We used the same EI scale, the WLEIS developed Measures of Sample 2
by Wong and Law (2002), as that used in Study 1. Coefficient alphas for
the EI dimensions of Self-Emotions Appraisal, Others-Emotions Appraisal, EI and personality. We used the same EI scale, the WLEIS developed
Regulation of Emotion, and Use of Emotion, as assessed by parents, were by Wong and Law (2002), as in Sample 1. Coefficient alphas for the four
.69, .84, .78, and .72, respectively. For the students self-ratings, the EI dimensions ranged from .76 to .90 for supervisory ratings, from .72 to
coefficient alphas for the above four EI dimensions were .69, .84, .81, and .89 for self-ratings, and from .77 to .91 for peer ratings. For personality
.72, respectively. For the personality dimensions, we used the 80-item Big dimensions, we used the 80-item Big Five personality measure introduced
Five personality measure of McCrae and Costa (1987). Coefficient alphas by McCrae and Costa (1987). Only self-ratings of the Big Five personality
for parents ratings of Neuroticism, Extraversion, Openness, Agreeable- traits were available. Coefficient alphas for Neuroticism, Extraversion,
ness, and Conscientiousness were .79, .83, .85, .85, and .90, respectively. Openness, Agreeableness, and Conscientiousness were .79, .80, .87, .85,
The coefficient alphas for the students self-ratings were .77, .80, .82, .83, and .90, respectively.
and .86, respectively. Task performance. Task performance was assessed by supervisors,
Life satisfaction. Life satisfaction was measured by four items that subordinates (self), and peers with three items derived from a general
were adopted from the life satisfaction scale developed by A. Campbell, performance measure (Farh & Cheng, 1997). A sample item is My
Converse, and Rodgers (1976). The coefficient alpha for these four items performance is very good as compared with other employees on the same
was .75.
CONSTRUCT AND CRITERION VALIDITY OF EI 491

Table 4 includes 15 items that capture two dimensions, interpersonal facilitation


Results of Regression Analyses on Life Satisfaction After and job dedication. Seven items are used to capture interpersonal facilita-
Controlling for Personality in Study 1 tion and 8 items are used to assess job dedication. The coefficient alphas
for supervisory, peer, and self-ratings of interpersonal facilitation were .82,
Sample 1 Sample 2 .86, and .85, respectively. For job dedication, coefficient alphas for the
three rating sources were .85, .89, and .83, respectively. Because the rater
Variables entered R2 AR2 R2 AR2 and the ratee varied when different respondents (supervisor, self, and peer)
were responding to the task and contextual performance items, the leading
Step 1 words in the questionnaire were adjusted accordingly.
NEURO .32** .36**
Control variables. Loyalty to supervisor, trust in supervisor, and the
EXTRA .24* .13*
OPEN .13** .03 demographic characteristics of the subordinates were used in the regression
AGREE .08* .25** analyses to control for variables that might affect job performance. We
CON .15* .354** .01 .203** used a 9-item scale to measure loyalty to supervisor, which was developed
Step 2 by Becker, Billings, Eveleth, and Gilbert (1996). The coefficient alpha was
WLEIS1 .09 .10 .89. A 3-item scale to measure trust in supervisor (Chen, 1997) was used.
and is not to be disseminated broadly.

WLEIS2 .04 .09 The coefficient alpha was .84. Age, gender (coded as 2 for women and 1
of its allied publishers.

WLEIS3 .19** .23** for men), education (months of education after graduation from high
WLEIS4 .00 .406** .052** .01 .265** .062** school), and tenure (months of experience with the immediate supervisor)
were the other demographic control variables.
Note. The regression weights of the control variables are the weights at
Step 1 only; the beta weights for the other variables are the final regression
weights after these variables were added to the model. N = 418 for Sample Analysis and Results
1; N = 314 for Sample 2. NEURO = Neuroticism; EXTRA = Extraver-
sion; OPEN = Openness; AGREE = Agreeableness; CON = Conscien- MTMM Approach
tiousness; WLEIS1, WLEIS2, WLEIS3, and WLEIS4 = the four dimen-
sions of the Wong and Law Emotional Intelligence Scale. Sample 1. Table 5 shows the descriptive statistics and results
p < .10. * p < .05. ** p < .01.
This document is copyrighted by the American Psychological Association

of our MTMM analyses of Sample 1. First, the reliabilities are


shown in parentheses on the main diagonal. EI was reliably mea-
job. The coefficient alphas for the three items, as reported by supervisor,
sured by the two methods of self-rating and parent rating. The
of the individual

peer, and self, were .80, .82, and .82, respectively. This variable is labeled coefficient alphas were .78 and .81, respectively. Acceptable reli-
task performance below. ability was found with the Big Five personality measures rated by
Contextual performance. Contextual performance was measured with the students (self) as well as by their parents. The coefficient
the scale developed by Van Scotter and Motowidlo (1996). The scale alphas ranged from .77 to .90. Second, the values of the
This article is intended solely for the personal

Table 5
Results of the 6 Traits 2 Methods MTMM Analyses in Sample 1 of Study 2

Self-rating Parent rating

Method and trait M SD EI NEURO EXTRA OPEN ANT CON EI NEURO EXTRA OPEN ANT CON

Self-rating
EI 3.60 0.46 (.78)
NEURO 3.67 0.93 .39 (.77)
EXTRA 4.74 0.94 .15 .08 (.80)
OPEN 4.93 0.95 .30 .12 .45 (.82)
ANT 5.21 0.74 .26 .36 .29 .14 (.83)
CON 5.04 0.77 .55 .46 .10 .27 .47 (.86)
Parent rating
EI 3.57 0.45 .28 .12 .00 .01 .02 .22 (.81)
NEURO 3.56 0.98 .18 .34 .04 .02 .18 .20 .30 (.79)
EXTRA 4.65 1.04 .06 .02 .37 .21 .02 .02 .00 .08 (.83)
OPEN 4.28 1.10 .15 .04 .14 .32 .10 .08 .15 .08 .55 (.85)
ANT 5.34 0.88 .07 .14 .01 .02 .20 .14 .16 .16 .28 .09 (.85)
CON 5.13 0.95 .17 .11 .13 .02 .05 .34 .42 .21 .11 .24 .58 (.90)

Note. The six traits are emotional intelligence (EI), neuroticism (NEURO), extraversion (EXTRA), openness (OPEN), antagonism (ANT), and
conscientiousness (CON). The two methods are self-rating and parent rating. The numbers on the diagonal are the coefficient alphas. Numbers in italics
are in the heterotraitmonomethod analyses; numbers underlined are in the heterotrait heteromethod analyses; numbers in bold are the results of the
monotrait heteromethod analyses. MMTM = multitraitmultimethod. Numbers in solid triangles are heterotraitmonomethod correlations; numbers in
dotted triangle are heterotrait heteromethod correlations.
492 LAW, WONG, AND SONG

heterotraitmonomethod triangles are shown in italics. Third, the Model B, which included only six traits, had a model chi-square of
values of the heterotrait heteromethod triangle are underlined. 8,449.92, a CFI of .68, and a TLI of .66. The changes in the model
Finally, the values presented in bold type are the monotrait chi-square of the two-method/six-trait model over these two-trait-
heteromethod values. only and the method-only models were both significant. Using a
Our results from Sample 1 concur with the requirements set by variance partitioning approach, we found that the six traits ex-
D. T. Campbell and Fiske (1959). First, the coefficients on the plained 41.49% of the total observed variance, the two methods
reliability diagonal (numbers in parentheses) are the highest in the explained 22.26% of the observed variance, and the remaining
matrix. Second, the coefficients on the validity diagonals (numbers 36.26% was attributed to random error.
in bold) are significantly different from zero and are high enough Sample 2. Table 6 shows the descriptive statistics and results
to warrant further investigation. The correlation between self- of our MTMM analyses of Sample 2. These results are very similar
rating EI and parent-rating EI was .28. Similar correlations for the to the results from Sample 1. First, the coefficients on the reliabil-
Big Five personality dimensions ranged from .20 to .37. The data ity diagonal (numbers in parentheses) are consistently the highest
showed discriminant validity, which requires that the validity in the matrix. Second, all of the coefficients in the validity diag-
coefficient be higher than all of the values in the column and row onals (numbers in bold) are significantly different from zero and
and is not to be disseminated broadly.

in which it is located. The correlation between parents EI ratings


of its allied publishers.

high enough to warrant further investigation. This indicates high


and students EI ratings of .28 seems low. One possible reason is convergent validity. Third, the correlations between self-rating,
that these are high school students, who may not yet have clear and peer-rating, and supervisory-rating of EI were .41, .38, and .34,
stable understandings of their own EI. Meanwhile, some Chinese respectively.
parents are quite protective of their children. In the minds of these The CFA results of the MTMM in Sample 2 were very similar
parents, their kids are always kids and the EI of their kids is always to those of Sample 1. In Sample 1, we needed to create three
quite low. The first reason implies that parents EI ratings may be random indicators for each construct because there were only two
more accurate than students self-ratings. The second reason im- sources for each trait. Because we had three methods (i.e., self,
plies that parents ratings of the EI of their kids are biased because peer, and supervisor) and four traits (EI, task performance, inter-
This document is copyrighted by the American Psychological Association

some parents would underestimate the EI of their kids but some personal facilitation, and job dedication) in Sample 2, we used a
may not. Without additional information, both arguments seem to single aggregate of each construct to simplify the model. In other
be plausible. words, the CFA consisted of 12 indicators from a combination of
of the individual

Fourth, some validity coefficients (numbers in bold) are smaller four latent traits and three latent methods. This three-method/four-
than some coefficients in the heterotraitmonomethod (numbers in trait model had a model y2(33, N = 274) = 79.85. The CFI and the
italic) triangles, which may imply that we have a common method TLI of this model were .98 and .95, respectively. Seventy percent
variance problem in the data set. However, these results should not of the total observed variance was attributed to the three methods,
be a worry. D. T. Campbell and Fiske (1959) commented that this 25% was attributed to the four traits, and 5% was attributed to
was a frequent phenomenon in research on individual differences. random error. The alternative model, with three methods only, had
y2(51, N = 274) = 283.91, a CFI of .88, and a TLI of .84. The
This article is intended solely for the personal

Finally, we see the same pattern of interrelationships among dif-


ferent traits in the two heterotraitmonomethod triangles and the alternative model, with four traits only, had y2(48, N = 274) =
heterotrait heteromethod triangle. 932.06, a CFI of .53, and a TLI of .36. Therefore, results from
Although D. T. Campbell and Fiskes (1959) recommended Sample 1 confirmed that the variance attributed to the traits was a
analyses are standard, they were not sufficiently precise or nor- major component of the total observed variance, whereas method
mative to evaluate the goodness of fit of the data with respect to (source) was the major component of total observed variance in
the prescribed model. Therefore, we used the factor analysis ap- Sample 2.
proach of the MTMM and ran a CFA with six traits (EI and the Big
Five) and two methods (self-rating and parent rating) on the data.
As in Study 1, we randomly grouped the items to form three
Hierarchical Linear Regression
indicators for each of the 12 constructs (six traits, each with two
rating sources). In the CFA, we specified a latent trait factor for Sample 1. To demonstrate the predictive validity of the EI
each of the six traits from the two sources (self-rating and parent construct, we ran a hierarchical linear regression of the parents EI
rating). We also prescribed two method factors for all the traits that ratings on the students ratings of powerlessness and life satisfac-
were assessed by self-rating and parent rating. However, because tion. After entering the control variables as the first step, we
we had only two methods (sources), the model was not identified. entered the Big Five personality dimensions as the second set of
To solve this problem, we followed Marsh and Hocevars (1983) control variables. We then entered the parents rating of EI as the
approach to obtain a priori estimates of the error variances of all final step. Table 7 shows the results of the hierarchical regression
the indicators. We conducted an exploratory factor analysis on all analyses. The parents ratings of EI were a significant predictor of
36 indicators and used one minus the estimated communality for the students life satisfaction, after controlling for demographic
each indicator as the estimate of the error variance of each variables and the Big Five personality dimensions ( = .16, p <
indicator. .05; AR2 = .02, p < .01). Hypothesis 2 was supported by the data.
Results of the CFA with EI, the Big Five personality dimen- It was also a significant predictor of students perceptions of
sions, and the two method factors of self and parent were encour- powerlessness over and above that of the Big Five personality
aging. For this two-method/six-trait, y2(649, N = 405) = dimensions ( = .17, p < .05; AR2 = .02, p < .05). Hypothesis
3,906.87. The CFI was .87 and the TLI was .85. We also tested two 3 was supported by the data.
alternative models. Model A, which included only two methods, Sample 2. In Sample 2, we had three dependent variables in
had y2(702, N = 405) = 15,048.18, a CFI of .41, and a TLI of .38. the hierarchical regression analyses: task performance, interper-
CONSTRUCT AND CRITERION VALIDITY OF EI 493

Table 6
Results of the 4 Traits 3 Methods MTMM Analyses in Sample 2 of Study 2

Self-rating Peer rating Supervisory rating

Method and trait M SD EI Task IF JD EI Task IF JD EI Task IF JD

Self-rating
EI 3.90 .46 (.89)
Task 3.90 .58 .54 (.79)
IF 4.06 .45 .65 .47 (.82)
JD 3.99 .48 .67 .58 .78 (.85)
Peer rating
EI 3.75 .47 .41 .25 .34 .38 (.89)
and is not to be disseminated broadly.
of its allied publishers.

Task 3.87 .63 .15 .39 .18 .29 .49 (.82)


IF 3.92 .52 .27 .17 .38 .34 .68 .46 (.86)
JD 3.87 .57 .17 .18 .29 .33 .67 .62 .78 (.89)
Supervisory rating
EI 3.65 .48 .38 .32 .19 .32 .34 .24 .15 .16 (.88)
Task 3.75 .66 .18 .31 .17 .30 .32 .38 .20 .27 .51 (.82)
IF 3.70 .53 .20 .19 .26 .27 .38 .26 .39 .31 .57 .44 (.85)
JD 3.71 .51 .23 .27 .26 .39 .45 .35 .32 .37 .58 .63 .69 (.83)
This document is copyrighted by the American Psychological Association

Note. Numbers in the triangles are heterotraitmonomethod correlations. Numbers in parentheses on the diagonal are the coefficient alphas. Correlations
among different raters of the same constructs (interrater reliability) are marked in bold. EI = emotional intelligence; Task = task performance; IF =
interpersonal facilitation; JD = job dedication.
of the individual

sonal facilitation, and job dedication. We entered the demographic Table 8 (Column M4A) shows that the peer rating of EI had
variables as controls (age, gender, education, and tenure with significant predictive power on task performance after controlling
supervisor) in the first step of the regression. Loyalty for supervi- for demographics, loyalty for supervisor, trust in the supervisor,
This article is intended solely for the personal

sor and trust in the supervisor were entered in the second step, and and the Big Five personality dimensions ( = .42; AR2 = .17, p <
peer ratings or employees self-ratings of EI were added in the .01). Similar significant results were found for the two contextual
final step. Table 8 presents the results of these regression analyses. performance dimensions. When contextual performance was used

Table 7
Results of Regression Analyses of Parent Ratings of EI on Academic Results, Life Satisfaction,
and Powerlessness

Life satisfaction Powerlessness

Variables entered Model 1 Model 2 Model 3 Model 1 Model 2 Model 3

Gender .05 .04 .04 .08 .09 .09


Working mother .07 .09 .09 .02 .01 .01
Fathers education .11 .08 .08 .08 .05 .05
Mothers education .06 .01 .01 .07 .11 .11
Neuroticism .17** .14* .14* .10
Extraversion .14 .14* .09 .08
Openness .10 .08 .25** .23**
Agreeableness .21** .22** .06 .07
Conscientiousness .03 .10 .07 .01
EI .16* .17*

R2 .04* .16** .18** .01 .10** .12**


AR2 .13** .02** .09** .02*
dfs 4, 284 5, 279 1, 278 4, 284 5, 279 1, 278
N 388 387

Note. The Big Five factors and EI are parents ratings. EI = emotional intelligence.
p < .10. * p < .05. ** p < .01.
494 LAW, WONG, AND SONG

Table 8
Results of Regression Analyses of Peer Ratings and Self-Ratings of EI on Job Performance

Task performance Interpersonal facilitation Job dedication

Variables entered M1 M2 M3 M4A M4B M1 M2 M3 M4A M4B M1 M2 M3 M4A M4B

Age .00 .05 .05 .04 .05 .11 .13 .13 .11 .13 .02 .05 .05 .02 .04
Education .09 .07 .07 .06 .06 .08 .07 .08 .08 .07 .06 .05 .07 .07 .06
Tenure with supervisor .16 .12 .13 .15* .12 .06 .07 .07 .03 .07 .02 .00 .00 .03 .01
Gender .13 .14 .14 .15* .13 .05 .03 .02 .01 .03 .10 .12 .14 .14* .12
NEURO (self) .03 .01 .02 .03 .04 .09 .07 .12 .08 .12 .11 .16
EXTRA (self) .04 .04 .00 .05 .01 .01 .04 .01 .05 .04 .01 .05
OPEN (self) .06 .06 .04 .03 .01 .00 .00 .03 .01 .01 .02 .06
AGREE (self) .26* .26** .27** .26* .12 .13 .13 .12 .17 .18 .18* .17
CON (self) .07 .06 .10 .06 .01 .01 .03 .01 .05 .07 .02 .06
and is not to be disseminated broadly.

Loyalty .12 .06 .10 .17 .13 .15 .22* .16 .19*
of its allied publishers.

Trust .03 .07 .06 .10 .15 .15 .19* .25** .25**
EI (peer) .42** .44** .50**
EI (self) .17* .26** .31**

AR2 .05 .05 .01 .17** .03* .02 .01 .02 .18** .06** .01 .04 .04 .24** .08**
AF 2.06 1.81 .88 35.97** 4.50* .91 .39 1.56 36.11** 9.71** .58 1.19 3.05 54.55** 15.29**
dfs 4, 163 5, 158 2, 156 1, 155 1, 155 4, 161 5, 156 2, 154 1, 153 1, 153 4, 163 5, 158 2, 156 1, 155 1, 155

Note. N = 168. Model 1 (M1) has only the demographic variables as predictors; Model 2 (M2) has the demographic variables plus the Big Five personality
dimensions; Model 3 (M3) adds in loyalty to supervisor (Loyalty) and trust in supervisor (Trust) as predictors; Model 4A (M 4A) has all predictors in Model
3 plus peer rating of EI; Model 4B (M4B) has all predictors in Model 3 plus self-rating of EI. NEURO = Neuroticism; EXTRA = Extraversion; OPEN =
This document is copyrighted by the American Psychological Association

Openness; AGREE = Agreeableness; CON = Conscientiousness; EI = emotional intelligence.


p < .10. * p < .05. ** p < .01.
of the individual

as the dependent variable, the peer rating of EI was a significant ses. Parents ratings of EI were also shown to account for incremental
predictor of interpersonal facilitation ( = .44; AR2 = .18, p < variance in life satisfaction and powerlessness beyond the Big Five
.01) and job dedication ( = .50; AR2 = .24, p < .01). Table 8 personality dimensions in a student sample. Peers ratings of EI were
(Column M4B) shows the results when employees self-ratings of EI found to be predictive of supervisory ratings of in-role and extrarole
were used as predictors of the supervisors rating of performance. performance in an employee sample.
Employees self-ratings of EI were a significant predictor of task There are at least three important implications arising from the
This article is intended solely for the personal

performance ( = .17; AR2 = .03, p < .05), interpersonal facilitation results of this article. First, if EI does indeed measure emotion-related
( = .26; AR2 = .06, p < .01), and job dedication ( = .31; AR2 = abilities that are distinct from personality traits, we certainly advocate
.08, p < .01). Hypothesis 4 was therefore supported by the data. continued research on EI and the development of scales that do not
rely on self-reports. Furthermore, given the confusion over the defi-
Discussion nition and domain of the EI construct in the past, we believe that it
In this article, we reviewed the definition and domain of the EI would be beneficial to EI researchers to adopt a mutually acceptable
construct and argued that it should be conceptually distinct from definition of the construct and to develop more standardized measures
traditional personality dimensions. We then used a two-study/four- according to this definition. As reviewed in this article, the four-
sample design to investigate the validity and utility of the EI dimensional definition adopted here appears to be a reasonable direc-
construct. We followed Davies et al.s (1998) work and other tion for future EI research. More research should be conducted ac-
recent work on EI (e.g., George, 2000; Mayer & Salovey, 1997; cording to this framework to avoid further confusion.
Mayer et al., 2000a) in defining EI as a four-dimensional construct, Second, our two studies provided evidence to support that EI is
comprising the ability to understand ones own and others emo- related to but distinct from the Big Five personality dimensions.
tions, to regulate ones emotions, and to use ones emotions. This Furthermore, Sample 2 of Study 2 showed that EI might be a good
definition of EI as a set of abilities conceptually distinguishes it predictor of job performance. After controlling for relevant variables
from personality traits, which are behavioral preferences. and the Big Five personality dimensions, EI still accounted for more
On top of this conceptual argument, we dealt with the empirical than 10% of the variance in in-role and extrarole performance when
conclusions of Davies et al. (1998) by using CFAs to investigate peer ratings of EI were used. It should be noted that neither EI nor job
various EI-related scales and a newly developed EI scale (the performance was assessed by the employees themselves. The results
WLEIS) on the basis of the four-dimensional view in Study 1. We are, therefore, not confounded by self-reporting. Given these initial
replicated Davies et al.s results from earlier EI-related scales, but positive results, researchers are encouraged to develop more rigorous
we further found that the WLEIS captured a construct that may be non-self-report measures of EI. One possible direction would be the
distinct from the Big Five personality dimensions. A CFA using the development of forced-choice EI questions, in which participants are
data from a second sample in Study 1 replicated the structural dis- asked to select the response that most closely represents their EI level,
tinctiveness of the EI construct from the Big Five personality dimen- rather than evaluating their own abilities directly.
sions. In Study 2, we obtained data on others ratings of EI to show Third, the criterion variables examined in this study included life
the convergent and discriminant validity of EI using MTMM analy- satisfaction, the feeling of powerlessness, and job performance. Con-
CONSTRUCT AND CRITERION VALIDITY OF EI 495

ceptually, EI may be more important for criterion variables such as Bagby, R. M., Taylor, G. J., & Parker, J. D. (1994). The twenty-item
psychological well-being and occupational stress experienced by Toronto Alexithyma Scale: II. Convergent, discriminant, and concurrent
workers. Future research should investigate the relationship between validity. Journal of Psychosomatic Research, 38, 33 40.
EI and criterion variables other than job performance. Furthermore, if Bagozzi, R. P., & Yi, Y. (1990). Assessing method variance in multitrait
EI is related to job performance and other important variables in the multimethod matrices: The case of self-reported affect and perceptions
at work. Journal of Applied Psychology, 75, 547.
workplace, training in EI may be as important as selecting applicants
Bar-On, R. (1997). Bar-On Emotional Quotient Inventory: A measure of
with high EI. Future research may attempt to develop effective EI
emotional intelligence. Toronto, Canada: Multi-Health Systems.
training programs. The predictive power of EI on job performance as Becker, T. E., Billings, R. S., Eveleth, D. M., & Gilbert, N. L. (1996). Foci
compared with general mental abilities could also be an interesting and bases for employee commitment: Implications for job performance.
research direction. Although we have some evidence that EI can Academy of Management Journal, 39(2), 464 482.
predict performance above and beyond general mental abilities, past Blau, P. M. (1964). Exchange and power in social life. New York: Wiley.
studies have used self-report measures of EI. It would be interesting Brislin, R. W. (1980). Translation and content analysis of oral and written
to compare the predictive validity of EI and general mental abilities material. In H. C. Triandis & J. W. Berry (Eds.), Handbook of Cross-
when EI is assessed by aptitude tests. cultural Psychology: Vol 2. Methodology (pp. 349 444). Boston: Allyn
and is not to be disseminated broadly.

& Bacon.
of its allied publishers.

There are, however, two limitations of this EI project. First,


although we have provided the validity of non-self-report assess- Campbell, A., Converse, P. E., & Rodgers, W. L. (1976). The quality of
ment of EI in Study 2, tests of EI might be important because EI American life: Perceptions, evaluation and satisfaction. New York:
Russell Sage Foundation.
is defined as an ability facet. Traditionally, abilities are measured
Campbell, D. T., & Fiske, D. W. (1959). Convergent and discriminant
by tests as opposed to self-report measures. The recent develop-
validation by the multitraitmultimethod matrix. Psychological Bulletin,
ment of the MayerSaloveyCaruso Emotional Intelligence Test 56, 81105.
(Mayer et al., 2000b) may be one possible move in this direction. Chen, Z. Y. (1997). Loyalty to supervisor, organizational commitment and
The major contribution of our study was to provide some prelim- employee outcomes: A Chinese case. Unpublished doctoral dissertation,
inary evidence of the validity of EI. By helping to open up this line Hong Kong University of Science and Technology, Hong Kong, China.
This document is copyrighted by the American Psychological Association

of research, we hope that there might be development of rigorous Ciarrochi, J. V., Chan, A. Y. C., & Caputi, P. (2000). A critical evaluation
tests of EI in the future, as well as evidence of its relationship with of the emotional intelligence construct. Personality and Individual Dif-
other organizational constructs. ferences, 28(3), 539 561.
of the individual

Our second limitation is that all the data in this project were Costa, P. T., & McCrae, R. R. (1985). The NEO Personality Inventory
collected in Hong Kong and the Peoples Republic of China. Cross- manual. Odessa, FL: Psychological Assessment Resources.
cultural generalizability of the results may be a concern. We do not Davies, M., Stankov, L., & Roberts, R. D. (1998). Emotional intelligence:
In search of an elusive construct. Journal of Personality and Social
know whether EI varies across different cultures. However, when we
Psychology, 75, 989 1015.
go back to the EI literature, we do not find any discussion of EI across
Dulewicz, V., & Higgs, M. (2000). Emotional intelligence: A review and
cultural boundaries. Our position is that ones abilities to understand, evaluation study. Journal of Managerial Psychology, 15(4), 341372.
regulate, and use ones emotions in constructive ways are general
This article is intended solely for the personal

Farh, J. L., & Cheng, B. S. (1997). Modesty bias in self-ratings in Taiwan:


human abilities. There is no immediate evidence that the validity of Impact of item wording, modesty value, and self-esteem. Chinese Jour-
EI, as defined under our four-dimensional view, should vary across nal of Psychology, 39(2), 103118.
cultures. Whereas further studies may be needed to verify this posi- Fineman, S. (Ed.). (1993). Emotion in organizations. London: Sage.
tion, we take the general scientific attitude that psychological and Friedman, H. S., Prince, L. M., Riggio, R. E., & Dimatteo, M. R. (1980).
management phenomena are considered as universal unless there are Understanding and assessing nonverbal expressiveness: The Affective
theories or evidence showing their cross-cultural variations. Communication Test. Journal of Personality and Social Psychology, 39,
Although the EI construct may be universal, we agree that behav- 333351.
iors resulting from the EI of an individual may vary across cultures. Gardner, H. (1993). Multiple intelligences: The theory in practice. New
York: Basic Books.
For example, a nonreactive quiet response by the subordinate when
George, J. M. (2000). Emotions and leadership: The role of emotional
ones boss is making unreasonable demands may reflect high EI
intelligence. Human Relations, 53(8), 10271055.
among Chinese workers but probably not among non-Chinese work- Hochschild, A. R. (1983). The managed heart: Commercialization of
ers. In this respect, our use of self-report measures of EI may, in fact, human feeling. Berkeley: University of California Press.
be a plus because we asked respondents about their final judgment of Lam, L. T., & Kirby, S. L. (2002). Is emotional intelligence an advantage? An
the EI of the target person irrespective of the assessment clues or exploration of the impact of emotional and general intelligence on individ-
methods they would use. By doing so, we may be able to avoid some ual performance. The Journal of Social Psychology, 142(1), 133143.
cross-cultural differences in expressing emotions or diagnosing emo- Marsh, H. W., & Hocevar, D. (1983). Confirmatory factor analysis of
tions, because the assessors would be able to use the clues or methods multitraitmultimethod matrices. Journal of Educational Measurement,
that are appropriate for their specific culture. This issue may have to 20(3), 231248.
be considered when behavior- or outcome-oriented tests of EI, such as Mathieu, J. E., & Farr, J. L. (1991). Further evidence for the discriminant
the Mayer-Salovey-Caruso Emotional Intelligence Test, are used validity of measures of organizational commitment, job involvement,
across cultural boundaries. and job satisfaction. Journal of Applied Psychology, 76, 127133.
Mathieu, J. E., Hofmann, D. A., & Farr, J. L. (1993). Job perceptionjob
satisfaction relations: An empirical comparison of three competing theories.
References Organizational Behavior and Human Decision Processes, 56, 370 387.
Mayer, J. D., Caruso, D. R., & Salovey, P. (2000a). Emotional intelligence
Atwater, L., & Yammarino, F. (1992). Does self other agreement on meets traditional standards for an intelligence. Intelligence, 27(4), 267298.
leadership perceptions moderate the validity of leadership and perfor- Mayer, J. D., Caruso, D. R., & Salovey, P. (2000b). Selecting a measure of
mance predictions? Personnel Psychology, 45(1), 141164. emotional intelligence: The case for ability testing. In R. Bar-On &
496 LAW, WONG, AND SONG

J. D. A. Parker (Eds.), Handbook of emotional intelligence (pp. 320 Schutte, N. S., Malouff, J. M., Bobik, C., Coston, T. D., Greeson, C.,
342). San Francisco: Jossey-Bass. Jedlicka, C., et al. (2001). Emotional intelligence and interpersonal
Mayer, J. D., & Salovey, P. (1997). What is emotional intelligence? In P. relations. The Journal of Social Psychology, 141(4), 523536.
Salovey & D. Sluyter (Eds.), Emotional development and emotional Schutte, N. S., Malouff, J. M., Hall, L. E., Haggerty, D. J., Cooper, J. T.,
intelligence: Educational implications (pp. 334). New York: Basic Golden, C. J., & Dornheim, L. (1998). Development and validation of a
Books. measure of emotional intelligence. Personality and Individual Differ-
McCrae, R. R., & Costa, P. T., Jr. (1987). Validation of the five-factor ences, 25, 167177.
model of personality across instruments and observers. Journal of Per- Sosik, J. J., & Megerian, L. E. (1999). Understanding leader emotional
sonality and Social Psychology, 52, 8190. intelligence and performance: The role of self other agreement on
Mehrabian, A., & Epstein, N. (1970). A measure of emotional empathy. transformational leadership perceptions. Group and Organization Man-
Journal of Personality, 40, 525543. agement, 24(3), 367390.
Pearlin, L., & Schooler, C. (1978). The structure of coping. Journal of Stough, C., & Palmer, B. R. (2002, July). The development of a workplace
Health and Social Behavior, 19, 221. measure of emotional intelligenceThe Swinburne University Emotional
Petrides, K. V., & Furnham, A. (2000). On the dimensional structure of Intelligence Test (SUEIT). Paper presented at the third international con-
emotional intelligence. Personality and Individual Differences, 29, 313 ference on Emotions and Organizational Life, Gold Coast, Queensland,
and is not to be disseminated broadly.

320. Australia.
of its allied publishers.

Roger, D., & Najarian, B. (1989). The construction and validation of a new Thorndike, E. L. (1920, March). Intelligence and its uses. Harpers Mag-
scale for measuring emotion control. Personality and Individual Differ- azine, 140, 227235.
ences, 10, 845 853. Van Scotter, J. R., & Motowidlo, S. J. (1996). Interpersonal facilitation and
Roseman, I. J., Dhawan, N., Rettek, S. I., & Naidu, R. K. (1995). Cultural job dedication as separate facets of contextual peformance. Journal of
differences and cross-cultural similarities in appraisals and emotional Applied Psychology, 81, 525531.
responses. Journal of Cross-Cultural Psychology, 26(1), 23 48. Williams, L. J., & Brown, B. K. (1994). Method variance in organizational
Salovey, P., & Mayer, J. D. (1990). Emotional intelligence. Imagination, behavior and human resources research: Effects on correlations, path
Cognition and Personality, 9(3), 185211. coefficients, and hypothesis testing. Organizational Behavior And Hu-
Salovey, P., Mayer, J. D., Caruso, D., & Lopes, P. N. (in press). Measuring man Decision Processes, 57(2), 185.
emotional intelligence as a set of abilities with the MSCEIT. In S. J. Wong, C. S., & Law, K. S. (2002). The effects of leader and follower
This document is copyrighted by the American Psychological Association

Lopez & C. R. Snyder (Eds.), Handbook of positive psychology assess- emotional intelligence on performance and attitude: An exploratory
ment. Washington, DC: American Psychological Association. study. The Leadership Quarterly, 13, 243274.
Salovey, P., Mayer, J. D., Goldman, S. L., Turvey, C., & Palfai, T. (1995). Wong, C. S., Wong, P. M., & Law, K. S. (2002, July). The interaction
of the individual

Emotional attention, clarity, and repair: Exploring emotional intelligence effect of emotional intelligence and emotional labor on job satisfaction:
using the Trait Meta-Mood Scale. In J. W. Pennebaker (Ed.), Emotion, A test of Hollands classification of occupations in Hong Kong. Paper
disclosure, and health (pp. 125-154). Washington, DC: American Psy- presented at the third international conference on Emotions and Orga-
chological Association. nizational Life, Gold Coast, Queensland, Australia.
This article is intended solely for the personal

Appendix

Development Process and Items in the the Wong and Law EI Scale (WLEIS)

Items in the WLEIS were generated, tested and cross-validated through 6. I am a good observer of others emotions.
a rigorous development process. The process started with three groups of 7. I am sensitive to the feelings and emotions of others.
MBA and undergraduate students trained on the EI construct generating 8. I have good understanding of the emotions of people around me.
items according to the four-dimensional definition of EI as proposed by
Davies et al. (1998). Inappropriate items were deleted according to three Use of Emotion (UOE)
criteria, resulting in a total of 9 items for each dimension. Items were 9. I always set goals for myself and then try my best to achieve them.
further eliminated on the basis of their factor loadings on their respective 10. I always tell myself I am a competent person.
factors as well as their correlations with selected criterion variables using 11. I am a self-motivating person.
a sample of 189 university students. Sixteen items (four items per dimen- 12. I would always encourage myself to try my best.
sion) were selected as a result. This 16-item scale was cross-validated with
Regulation of Emotion (ROE)
three additional student samples (N1 = 72, N2 = 146, N3 = 110), 116
13. I am able to control my temper so that I can handle difficulties
nonteaching employees from a university, and 149 supervisorsubordinate
rationally.
dyads (60 middle- and upper-level managers). The entire development
14. I am quite capable of controlling my own emotions.
process is described in Wong and Law (2002) in detail.
15. I can always calm down quickly when I am very angry.
16. I have good control of my own emotions.
The 16 items in WLEIS are as follows:
Note. The 16 items in WLEIS are reprinted for The Leadership Quarterly,
Self-Emotions Appraisal (SEA) 13, C. S. Wong and K. S. Law, The effects of leader and follower
1. I have a good sense of why I have certain feelings most of the time. emotional intelligence on performance and attitude: An exploratory study,
2. I have good understanding of my own emotions. 243274, Copyright (2002), with permission from Elsevier.
3. I really understand what I feel.
4. I always know whether or not I am happy.
Received October 11, 2002
Others-Emotions Appraisal (OEA) Revision received July 16, 2003
5. I always know my friends emotions from their behavior. Accepted July 28, 2003

You might also like